Conjunction and linear non-separability effects in visual shape encoding

Four visual search experiments are reported which used simple 2D shapes varying on the global dimensions of aspect ratio/curvature or aspect ratio/tapering. Results indicate serial self-terminating search in all conditions. Most importantly, search rates are markedly modulated by the particular forms of structural relations existing between the targets and their distractors. Thus, single-feature targets with shape properties that are linearly separable from those of their distractors yield markedly faster search rates than linearly separable targets made of a conjunction of distractor features. In addition, linearly separable single-feature targets are searched at a much faster rate than single-feature targets which are not linearly separable. Follow-up experiments demonstrate that these conjunction and linear non-separability effects cannot be attributed to pairwise target-distractor discriminability differences across conditions. The main conclusions are that the shapes used are parsed according to elementary features in visual encoding, and that a linear discrimination mechanism is available which permits fast visual search rates if a single-feature target is linearly separable from its distractors.